Establishment And Clinical Verification Of A Mathematical Model For Predicting The Probability Of Malignancy Or Begin In Patients With Solitary Pulmonary Nodules

ObjectiveTo establish a mathematical model to estimate the probability of malignancy in patients with solitary pulmonary nodule,and to verify the accuracy of the model comparing with foreign models(Mayo model and VA model).MethodsA retrospective cohort study included 405 patients (220 males and 174 females) with definite pathological diagnosis of SPN from Jan 2008 to Sep 2013 (group A). Clinical and imaging features were analyzed, and a clinical prediction model was built with multivariate Logistic regression analysis. Other 168 SPN patients (group B) with definite pathological diagnosis from Oct 2013 to Sep 2015 were used to estimate the accuracy of the model, and the model was also compared with Mayo model and VA model (two classical clinical prediction models reported in the literature). Calibration of the 3 models (our model、Mayo model and VA model) was assessed by Hosmer-Lemeshow (H-L) test, and discrimination was tested by calculating the area under curve (AUC) after the receive operating characteristic (ROC) curve was drawn.ResultsLogistic regression analysis showed that eihgt clinical characteristics (age of patient, previous cancer history, family history of cancer, diameter, speculation, lobulation, border, pleural retraction sign) were independent predictors of malignancy in SPN patients (P<0.05). The sensitivity in group B was 94.7%, specificity 76.4%, positive predictive value 89.2%, and negative predictive value 87.5%. The H-L test showed good fitting in 3 models (P>0.05). The AUC our model was 0.809±0.017, and 0.764±0.033 in Mayo model,0.715±0.012 in VAmodel.Conclusion1.Eight clinical characteristics (age of patient, previous cancer history, family history of cancer, diameter, speculation, lobulation, border, pleural retraction sign) were independent predictors of malignancy in SPN patients.2.The clinical prediction model that we established by retrospective study is:P=ex*/ (1+ ex),X=-6.402+(0.062×age)+(1.547× previous cancer history)+(1.195×family history of cancer)+(0.327×diameter)+(1.040×speculation)+(0.716xlobulation)+(0.279x border)+(1.934xpleural retraction sign)3.The clinical prediction model that we established by retrospective study has a higher clinical value, and it’s predictive effect is better than the other clinical prediction model.